ClinicalTrials.gov MCP Server

An MCP server (Apify Standby mode) that lets an AI agent search ClinicalTrials.gov, the official US NIH/NLM clinical trial registry, directly from a tool call — no scraping, no dataset to download, no separate Actor run to poll. Point your MCP client at this Actor's Standby URL and call search_clinical_trials like any other tool.

$0.005 per answered call, flat. No Actor-start fee. No result, no charge.

Tool: search_clinical_trials

Argument	Type	Notes
condition	string, optional*	Medical condition, e.g. "asthma"
intervention	string, optional*	Drug, device or procedure, e.g. "semaglutide"
maxResults	number, optional	Default 10, max 1000

* At least one of condition or intervention is required. With neither, the registry API returns an arbitrary slice of all ~550k studies, which reads like a real answer to an agent but isn't one — so the tool refuses the call instead, and that refusal is not charged.

Returns a JSON array of studies, each with nctId, title, status (the registry's overallStatus, e.g. RECRUITING, COMPLETED, NOT_YET_RECRUITING) and a canonical url.

Example tool-call request

{"name": "search_clinical_trials",
 "arguments": {"intervention": "semaglutide", "maxResults": 2}}

Example output

[
  {
    "nctId": "NCT00813020",
    "title": "A Clinical Trial Investigating the Comparison of Three Different Concentrations of NN9535 in Healthy Male Subjects",
    "status": "COMPLETED",
    "url": "https://clinicaltrials.gov/study/NCT00813020"
  },
  {
    "nctId": "NCT05567796",
    "title": "A Research Study to See How Well CagriSema Helps People With Excess Body Weight Lose Weight",
    "status": "ACTIVE_NOT_RECRUITING",
    "url": "https://clinicaltrials.gov/study/NCT05567796"
  }
]

Live output, captured 2026-11-05 against the published Standby URL.

What it is for

• Patient/trial triage mid-conversation — an agent asked "are there recruiting trials for this condition?" answers in one tool call instead of running a batch scrape and waiting.
• Pharma and biotech competitive questions — "what trials involve this drug?" without standing up a pipeline.
• Grounding a medical answer in a citable registry record — every row carries its nctId and canonical ClinicalTrials.gov URL, so the agent can cite rather than assert.
• Cheap breadth checks before committing to a bulk pull — probe a condition for $0.005, then switch to the batch Actor below if the answer is worth 39 fields.

Pricing and how we compare

Pay-per-event, a flat $0.005 per call that returned at least one study. No start fee, no platform run to babysit, no per-study multiplier — a call returning 500 studies costs the same as one returning 1.

This is the only MCP server on the Apify Store dedicated to ClinicalTrials.gov (checked 2026-11-05 across the clinical trial, mcp server, medical mcp and pubmed queries — the 19 other ClinicalTrials.gov listings are all batch scrapers, a different product shape). Trial search does exist as one tool inside a broader compliance bundle, and that is the honest comparison to make. Prices read live from each Actor's own pricingInfos record on 2026-11-05, not from Store marketing copy:

Actor	Users (u30d)	Closest event	Price/call	vs. ours
fetchsmith/clinicaltrials-mcp-server	—	tool-call	$0.005	—
red.cars/healthcare-compliance-mcp	2 (1)	search_clinical_trials	$0.05 + $0.00005 start fee	10x dearer, plus a start fee we don't charge
red.cars/pharma-pipeline-intelligence-mcp	8 (1)	search_drug_pipeline	$0.05	10x dearer
themineworks/academic-research-mcp	33 (29)	search_pubmed	$0.0625 + $0.005 start fee	12.5x dearer, plus a start fee we don't charge

All three are genuinely broader than this Actor. healthcare-compliance-mcp exposes 9 tools over FDA MAUDE adverse events, 510(k) clearances and recalls — including a search_clinical_trials of its own at $0.05 and a get_trial_details at $0.03; pharma-pipeline-intelligence-mcp exposes 8 (competitive landscape, adverse-event signals, patent exclusivity); academic-research-mcp exposes 6 across OpenAlex/Crossref/arXiv/OpenCitations/PubMed. If you want a device-compliance workflow or a multi-source literature report, buy theirs — this Actor does not try to be that.

What this Actor is, that none of them are: the registry lookup on its own, at a tenth of the price, with no start fee and no charge when the search finds nothing. If your agent asks trial questions often, the per-call difference is the whole cost of the feature.

Source and limits

Backed by ClinicalTrials.gov's own public API v2 — no API key, no login, no proxy, no scraping, no ToS grey area. maxResults is clamped to 1000 because pageSize=1001 returns 200 with only the first 1000 rows honored, so a higher value would silently under-deliver. Results are the registry's own relevance order; this tool does not paginate beyond the first page, by design (an agent asking a live question wants the top matches, not a bulk export).

Same source as our clinicaltrials-scraper Actor. Use that one instead if you need the full 39-field dataset, site-level rows, eligibility criteria, sponsors, or a scheduled/watched pull; use this MCP server when an agent needs to ask a live question mid-conversation.

FAQ

Do I need an Anthropic/OpenAI key for this? No. An MCP server serves tools; your agent brings its own model. This Actor makes zero inference calls and has no LLM in it — you are paying for registry access, not for tokens.

What happens if ClinicalTrials.gov is down or slow? The tool returns an MCP error result saying so, and does not charge you. Requests time out at 20 seconds. Retry the call.

Am I charged for a search that finds nothing? No. A zero-match search is free, as is a call rejected for a bad or missing argument. You are only charged for a call that returned at least one study.

Which MCP clients work? Any client speaking Streamable HTTP MCP — Claude Desktop/Code, Cursor, Codex, and the Apify MCP tooling. Point it at this Actor's Standby URL with /mcp as the path and your Apify token.

Why only four fields? Because an agent reading a tool result pays for every token of it. nctId, title, status and url are what a live question needs, and the url is the escape hatch to the full record. If you need the other 35 fields in bulk, the batch Actor above is the cheaper way to get them.

Related guides

• fetchsmith.com/tools
• clinicaltrials-scraper — batch, 39 fields
• An MCP tool that refuses bad calls instead of guessing — and doesn't charge for the refusal
• The ClinicalTrials.gov API silently caps pageSize at 1000 — and its phase filter doesn't exist where you'd look for it
• A 4-day window on ClinicalTrials.gov returns more trials than a 21-day one